Download comparative imaging study using a multi-modality

COMPARATIVE IMAGING STUDY IN ULTRASOUND,
MRI, CT AND DSA USING A MULTI-MODALITY
RENAL ARTERY PHANTOM
DEIRDRE M. KING*, ANDREW J. FAGAN†, CARMEL M. MORAN‡, AND
JACINTA E. BROWNE*
*
Medical Ultrasound Physics and Technology Group, School of Physics, Dublin
Institute of Technology, Dublin 8, Ireland,
†Centre for Advanced Medical Imaging (CAMI), St James's Hospital, Dublin 8,
Ireland and
‡Department of Medical Physics, University of Edinburgh, Edinburgh, UK
Abstract
Purpose: A range of anatomically-realistic multi-modality renal artery phantoms
consisting of vessels with varying degrees of stenosis was developed and evaluated
using four imaging techniques currently used to detect renal artery stenosis (RAS).
The spatial resolution required to visualize vascular geometry and the velocity
detection performance required to adequately characterize blood flow in patients
suffering from RAS is currently ill-defined, with the result that no one imaging
modality has emerged as a gold standard technique for screening for this disease.
Methods: The phantoms, which contained a range of stenosis values (0, 30, 50, 70
and 85%), were designed for use with ultrasound, magnetic resonance imaging, X-ray
computed tomography and X-ray digital subtraction angiography. The construction
materials used were optimized with respect to their ultrasonic speed of sound and
attenuation coefficient, MR relaxometry (T1, T2) properties, and Hounsfield number /
X-ray attenuation coefficient, with a design capable of tolerating high-pressure
pulsatile flow.
Fiducial targets, incorporated into the phantoms to allow for
1
registration of images among modalities, were chosen to minimize geometric
distortions.
Results: High quality distortion-free images of the phantoms with good contrast
between vessel lumen, fiducial markers and background tissue to visualize all
stenoses were obtained with each modality. Quantitative assessments of the grade of
stenosis
revealed
significant
discrepancies
between
modalities,
with
each
underestimating the stenosis severity for the higher-stenosed phantoms (70% and
85%) by up to 14%, with the greatest discrepancy attributable to DSA.
Conclusions: The design and construction of a range of anatomically-realistic renal
artery phantoms containing varying degrees of stenosis is described. Images obtained
using the main four diagnostic techniques used to detect RAS were free from artifacts
and exhibited adequate contrast to allow for quantitative measurements of the degree
of stenosis in each phantom. Such multi-modality phantoms may prove useful in
evaluating current and emerging US, MRI, CT and DSA technology.
Key words – Ultrasound, multi-modality, phantom, in vitro experimentation
2
INTRODUCTION
5
The renal system plays an important role in maintaining the health of individuals, for
example through the removal of waste products from the blood and the controlling of
blood pressure and the manufacture of red blood cells. A renal artery stenosis (RAS)
can result in clinical syndromes such as renovascular hypertension, ischemic
10
nephropathy and pulmonary edema [1, 2], and is associated with an increased risk of
cardiovascular mortality. The vast majority of stenoses are caused by atherosclerosis,
a common progressive problem that increases in prevalence with age [2] and it is
recognized as an important cause of secondary hypertension which afflicts
approximately 60 million people in the USA alone [3, 4]. Current evidence also
15
suggests that the presence of atherosclerotic plaque in the renal artery is indicative of
the presence of plaque at other sites in the body such as the carotid artery or the
coronary artery, which further adds to the cardiovascular risk and long-term prognosis
for patients with RAS [5]. Thus, there is a considerable advantage in detecting RAS
at the earliest possible stage of the disease, in order to maximize the therapeutic
20
outcome following drug administration or renal vascularisation.
Consequently, there is much clinical interest in the initiation process and progression
of atherosclerotic disease in the renal artery, the relationship between stenosis,
25
hypertension and end-stage renal failure, and whether more accurate stenosis
assessment techniques would improve patient outcome.
The medical imaging
techniques currently available to investigate this condition are ultrasound (US), X-ray
computed tomography (CT), magnetic resonance imaging (MRI) and X-ray digital
subtraction angiography (DSA). However, criteria for selecting the most suitable
3
30
technique for screening patients with suspicion of RAS are currently ill-defined, in
terms of the required spatial resolution and velocity detection performance for
quantification of the vascular geometry and the characterization of the blood flow.
Furthermore, only a limited quantitative assessment of the detection capabilities of
these imaging modalities has been carried out to establish the optimum modality or
35
combination of modalities for the diagnosis of renal artery stenosis, in large part due
to the lack of a suitable multimodality phantom.
Before a medical imaging technique can be introduced to routine clinical use as a new
imaging alternative to the previous gold standard, it must first be validated in order to
40
determine its accuracy, a process usually carried out via expensive and lengthy
clinical trials [6]. Such clinical trials often have complicated ethical issues associated
with them, experience difficulties recruiting a sufficiently large number of patients
which would be required to achieve statistical significance between the performance
of the new and existing gold standard techniques. Furthermore, the clinical truth is
45
seldom known, so it is difficult to assess the absolute accuracy of each of these
imaging techniques.
However, the validation procedures for new imaging techniques can also include
in vitro flow phantom experimentation [7-10] wherein a valid comparative study
50
could be conducted by imaging a standardized vascular phantom with each modality,
with the phantom ideally containing fiducial markers to aid image registration. Four
such multi-modality vascular phantoms have been developed and described in the
literature, each of which contained fiducial markers to aid image registration [7-9,
11]. Frayne et al. developed a carotid vascular phantom for comparative studies of X-
4
55
ray DSA, US and MRI [9]. However, the vessel mimicking material, polyester resin,
used for the vessel wall in this phantom was found to be only suitable for X-ray
imaging, with artifacts occurring in both US and MR images. Furthermore, the
authors used lead core fishing line (0.8 mm outer diameter with 0.2 mm diameter lead
core) as the fiducial marking system; this marker system was only clearly detectable
60
in the X-ray images. The approach adopted by Dabrowski et al. used a formaldehydefixed section of a real human iliac artery cannulated onto two acrylic tubes to mimic
the vessel wall, and three imaging modalities (DSA, CT, and 3-D B-mode US) were
evaluated by comparing images of this phantom [8]. Although the use of ex vivo
human vessels assured accurate geometric and tissue properties for this phantom, the
65
use of ex vivo human tissue inherently reduces the lifespan of such phantoms and
further renders it impossible to carry out longitudinal or inter-institutional studies. An
additional problem that was encountered with this phantom stemmed from the use of
small stainless steel ball bearings for the fiducial markers, which introduced
significant artifacts in both the US and CT images and would further pose
70
considerable problems for MR imaging. Cloutier et al. developed a multi-modality
phantom containing a straight lumen with a double stenosis (70% and 85%) [7];
however, the low acoustic velocity compared to in vivo tissue of the agar-oil mixture
used in this phantom caused geometric distortion in the ultrasound images, while the
latex used to mimic the vessel wall caused a slight reduction in the measurement of
75
the vessel lumen diameter.
Although this phantom incorporated two grades of
stenosis, the positioning of the stenoses along a straight lumen did not replicate the
complexity of a clinical presentation of RAS, where stenoses typically occur near
bends in vessels where turbulent flow effects lead to mechanical wall stresses and
regions of slow flow conducive to plaque formation/deposition.
5
More recently,
80
Allard et al. incorporated the use of an alternative to low melting point metal, Isomalt
(which is a commercial sugar alcohol), for the fabrication of the vessel geometries
[11]. This phantom also used the same layout of fiducial markers as Cloutier et al.,
however volume shrinkage was observed during the cooling phase for this material
which may have affected the final lumen dimensions. There is a clear need for the
85
development of an anatomically-realistic phantom which can be easily and
reproducibly manufactured and imaged in a distortion-less manner with each
modality, while also incorporating features of disease progression such as vessel
stenoses.
90
An alternative approach to building physical models is to use the technique of rapid
prototyping, wherein a structure of interest is constructed layer by layer from a 3D
computer model of the structure. Using rapid prototyping, it is possible to rapidly and
reproducibly produce models with complex shapes, which makes the technique
particularly attractive for developing anatomically-realistic phantoms. In the current
95
work, the renal artery morphology was modeled using a 3D dataset from a CT scan of
a patient imported into a computer-aided design (CAD) software package, from which
a physical model of the artery was produced using rapid prototyping. This model was
then used to produce anatomically-realistic renal artery flow phantoms with in-built
fiducial markers, using customized materials which optimally mimicked tissue
100
properties with respect to ultrasound, MRI and X-ray DSA/CT.
Attention was
focused on ensuring compatibility with each modality, from both a safety perspective
(for example, no ferromagnetic materials were used) and also the ability to produce
artifact-free images with adequate image contrast with each modality. In addition, a
range of stenosis values (0, 30, 50, 70, 85%) were introduced into the phantoms, in
6
105
order to investigate the usefulness of the phantoms in assessing the geometric
accuracy of each modality. The aim of the work was thus to develop a suitable RASmimicking phantom and demonstrate its potential in a comparative studies involving
US, MRI, CT and DSA.
110
METHODS AND MATERIALS
115
TMM characterization
An agar based tissue mimicking material (TMM) developed by Terlinck et al. (1998)
[12] was chosen for the phantoms, the constituents of which are give in Table 1, and
its manufacture is described elsewhere [13]. The acoustic properties, in particular
120
speed of sound and attenuation coefficient were verified using a Scanning Acoustic
Macroscope with a broadband transducer centered at 2.25 MHz (Ultrasonic Sciences
Limited, Fleet, UK) using the pulse-echo substitution technique [14]. To verify the
MR properties of the TMM, T1 and T2 relaxometry measurements were carried out
using a two-point ratio method with a mixed sequence implemented on a 3T Philips
125
Achieva MRI System (Philips Medical Systems, Netherlands), wherein inversion
recovery (IR) and spin echo (SE) pulse sequence modules were consecutively
combined in one cycle to provide a measure of T1 and T2. This technique is
insensitive to phase errors due to the use of a phase correction built into the sequence.
The SE and IR repetition time (TR) values used for the image acquisition were
130
6000 and 7000 ms respectively, chosen to allow for a full recovery of magnetisation
in each case, while the IR delay time was set at 800 ms. The T2 measurements were
carried out using a Carr-Purcell Meiboom-Gill sequence module with eight echoes
from 10 - 80 ms in increments of 10 ms. The Hounsfield (CT) number was
7
determined from an image of the TMM acquired using a 64-slice Sensation (Siemens
135
Medical Solutions, Karlsruhe, Germany), CT scanner at 120 kVp.
Although it was
not possible to accurately measure the X-ray attenuation coefficient of the TMM (as a
mono-energetic X-ray source was not available), a transmission comparison was
carried out between the TMM and solid water phantoms using a standard X-ray
imaging system (Digital Diagnostic system, Philips Medical Systems, The
140
Netherlands) and a radiation detector (Model 9010 Radiation Monitor Controller with
6cc ionisation chamber, Radcal Corp., USA).
Measurements were performed at
multiple kVp settings (50, 60, 81, 102 and 125 kVp) using TMM and water
thicknesses of 5 cm.
145
Table 1 Weight composition of the TMM
Component
Weight composition (%)
Distilled water
82.97
Glycerol
11.21
Silicon carbide (400 grain)
0.53
Aluminium oxide (3 μm)
0.94
Aluminium oxide (0.3 μm)
0.88
Benzalkoniumchloride
0.46
Agar
3.0
Phantom construction
150
To develop an anatomically-realistic renal artery flow phantom, a CAD-based model
of a healthy renal artery was developed from a clinical 3D CT dataset.
Ethics
approval was obtained from Dublin Institute of Technology’s Research Ethics
Committee. The CT images were imported into a commercial software package
(MIMICS 10, Materialise, Leuven, Belgium) which provided an interface for the
155
visualization and segmentation of CT images and the 3D rendering of a region of
8
interest (ROI). The computer model of the renal artery was exported from SOLID
WORKS into the Rapid Prototyping machine, a Z printer 310 Plus 3D printer
(Zcorporation, MA, USA), in the form of STL files from which a physical model of
the artery was produced by rapid prototyping. The CAD computer model used to
160
produce the healthy artery, which had a diameter of 6.8 mm, is presented in Figure 1.
The physical model thus produced was then used to create a master silicone negative
mould, into which a low melting point alloy (melting point: 47ºC, MCP 47, Mining
and Chemical Products Ltd., Northamptonshire, UK) was poured. Once the metal core
had set, it was removed from the silicone mould and positioned within the phantom
165
container and the liquid-phase TMM was poured around it and in turn allowed to set.
The metal core was subsequently removed through melting by heating the phantoms
to 50ºC ≤ T ≤ 60ºC, which is in excess of the melting temperature of the metal alloy.
In this way, it was possible to create a lumen in the TMM with an anatomicallyrealistic geometry.
170
Figure 1 Normal renal artery computer model
In addition to the healthy artery (i.e. with no stenosis), a range of other arteries were
produced with stenoses of 30, 50, 70 and 85%. These were produced by using a
175
series of symmetrical inserts for each grade of stenosis, each of which was 10 mm in
length (Figure 2). These inserts were placed in the master silicon negative mould
before the liquid metal alloy was poured, and hence the shape of the resultant metal
9
cores, which were used to create the vessel lumen, already contained the shape of the
different degrees of stenoses. To ensure the entire metal alloy was removed from the
180
stenosed phantoms, the metal core was wrapped in cling film prior to its placement in
the phantom container and the addition of the TMM. Thus, when the phantom was
heated, the entire liquefied metal alloy was removed together with the cling film [15].
Following discussions with an experienced vascular clinical measurement scientist
and a review of the literature, the position of the stenoses was chosen to lie
185
approximately 10 mm from aorta which corresponds to 10 mm from the inlet of the
model as it has been shown that more than 90% of stenoses are caused by
atherosclerosis, which typically occur at the renal artery origin [9].
Additional
information regarding the manufacture of these phantoms is provided elsewhere [15].
190
Figure 2 Removal inserts used to fabricate the diseased models of the renal artery.
Reproduced with permission [15]
195
In addition to providing a realistic comparison of the imaging modalities, it was
important to ensure that the image acquisition parameters used in each case reflected
those used clinically. Such parameters are often determined automatically by the
10
imaging systems, for example MRI uses automatic pre-scan calibration procedures to
determine detector coil loading, radiofrequency power and detector gain levels, while
200
X-ray DSA systems use feed-back mechanisms such as the automatic exposure
control (AEC) to determine X-ray energy and flux intensities required to optimize the
image quality. Phantom sizes with dimensions 115 mm x 79 mm x 92 mm
(L x W x H) presented a reasonable volume for imaging to the respective modalities
and also helped to avoid edge-effects when assessing the appearance of the renal
205
arteries in the images.
Glass beads (Crown Jewels, Dublin, Ireland) with a diameter of 4 mm were chosen as
suitable fiducial markers to facilitate image registration across the modalities. Using
sewing thread (diameter 0.8 mm), four rows of four beads were threaded through the
210
phantom, with two rows of beads located on either side of the renal artery lumen. On
each row the beads were separated from each other by 2.8 mm and secured with
adhesive. The bottom row was secured 57 mm from the top of the container and 20
mm from the edges of the container. The upper row of each pair was placed 8 mm
above the first and 10 mm from the edge of the container. This lateral displacement
215
of the top rows relative to the position of the lower rows was used to minimize any
obstruction of the lower row caused by shadowing in the US images deriving from the
upper row. The glass bead marker system is presented in Figure 3.
11
(a)
(b)
*
220
Figure 3 (a) Cross-sectional schematic diagram of the fiducial markers in the phantoms,
(b) Photograph showing the positioning of the glass bead markers before inclusion of the
metal artery model and background TMM, where * marks the location of the stenosis relative
to the fiducial markers.
225
Image acquisition of phantoms
The phantoms were imaged with standard clinical imaging systems, using basic
imaging protocols rather than specific clinical protocols or flow measurements, since
the aim of the study was to assess the visualization of the phantoms with each
imaging modality and determine the respective geometric accuracies. The phantoms
230
were imaged with no liquid in the vessels for CT and MRI, with water in the vessels
for US, and with a water-contrast agent mixture for DSA.
Longitudinal and
transverse B-mode ultrasound images were acquired with a HDI 3000 system (Philips
Medical Systems, Netherlands) using a linear array transducer (L12-5) at a nominal
frequency of 6 MHz. The images were optimized in each case with the output power
235
set at maximum, focal position at the centre of the vessel and a field of view of 8 cm.
12
3D T2-weighted MR images were acquired with a 3T Philips Achieva system (Philips
Medical Systems, Netherlands) using a turbo spin echo (TSE) technique, with the
phantoms placed inside an 8-channel phased array detector coil (Philips Head
SENSE 8).
240
Images were obtained using the following acquisition parameters:
resolution = 0.8 x 0.8 x 0.8 mm, field of view =150 x 141 x 130 mm, time to echo
(TE) = 380 ms, time to repetition (TR) = 2500 ms, Turbo factor = 120, number of
signal averages = 2, SENSE parallel imaging factor = 2. CT images were acquired
with a GE Discovery STE system (GE Healthcare, UK) which included a full GE
light speed helical 16-Slice CT using a slice thickness of 0.6 mm, a scan range of
245
120 mm, and a matrix size of 1024 x 1024. The current was 250 mA and peak
voltage was 120 kV. The computed tomography dose index (CTDI) was 35.52 mGy
and dose length product (DLP) was 572.2 mGy.cm. DSA images were acquired on a
Philips Integris Allura System, (Philips Medical Systems, Netherlands), with each
vascular phantom filled initially with water and then with a contrast solution of
250
Ultravist (Bayer Schering Pharma AG, Germany) at 240 mg ml-1. The concentration
of the contrast agent was carefully selected to mimic the image intensity enhancement
obtained clinically.
Projections were obtained parallel to the vessel using the
following parameters: source-to-detector distance = 67 cm, field-of-view = 23 cm,
table height 99 cm, current = 400 mA and peak voltage = 50 kV. Each image
255
sequence was captured at a frame rate of 6 frames per second with 4 ms of exposure
with a matrix size of 1024 x 1024.
Diameter measurements
To determine the spatial accuracy of each imaging modality, the diameter of the
260
vessel lumen was measured at three locations in the respective images of each
13
phantom (0, 30, 50, 70 and 85% stenosis); these locations are illustrated in Figure 4
(position A – before stenosis: 10 mm from inlet; position B – at centre of stenosis:
15 mm from inlet; and position C - after stenosis: 10 mm from bend).
These
measurements were compared with reference measurements of the diameter of the
265
metal model used to produce the lumen in each phantom, which were measured five
times at each location using a Vernier caliper. A one way ANOVA statistical test was
carried out to compare the vessel lumen diameters determined from each imaging
modality. The null hypothesis being tested was that there was no difference in the
dimensions of the models. If p < 0.05, the null hypothesis was rejected and there was
270
a statistical difference in the dimensions of the models. All statistical tests were
carried out using Minitab® Statistical Software (version 15).
Figure 4 Location of the reference measurements on the metal model at position A-before
stenosis: 10 mm from inlet; position B- at stenosis: 15 mm from inlet; and position C – after
275
stenosis: 10 mm from bend.
Transverse B-mode was used to determine the diameter of the vessels in the US
images. For MRI and CT images, the fiducial markers were used to identify and
orientate the plane views, and the coronal slices containing the mid-point of the vessel
280
for each phantom were selected. For DSA, the diameter was calculated from the
subtracted contrast images for each phantom.
All of the acquired images were
analyzed using a MATLAB (The Mathworks, USA) program developed in-house.
This program allowed for a line profile to be drawn across the lumen extending over
14
the vessel-lumen boundaries at each position, from which the vessel lumen diameter
285
could be determined by plotting the image intensity versus distance across each region
for all modalities; an example is illustrated in Figure 5 for a transverse B-mode US
image. A graphical interface in the MATLAB program was used to display the
intensity profile along the line of interest, allowing the user to specify the beginning
and end of the vessel. The peaks corresponding to the edges of the near and far walls
290
were located.
The distance in [mm] was calculated by measuring the known
dimensions of the phantom and converting the pixel distance in mm.
Figure 5 Calculation of the vessel diameter in a B-mode ultrasound image: (a) line profile
marked across the lumen, (b) plot of image intensity versus distance along the chosen line
295
profile. Reproduced with permission [15]
RESULTS
TMM Characterization
300
The measured TMM properties and average in vivo properties for soft tissue in the
vicinity of the renal artery are presented in Table 2. In general, the renal artery is
surrounded by fatty tissue which can typically be of the order of 1-2 cm thick,
increasing for obese patients. Beyond the fat layer, a range of other soft tissues are
found, for example muscle, pancreases, duodenum (around the right RA), together
305
with several blood vessels (e.g. renal vein, aorta, inferior vena cava). The TMM had
15
similar acoustic properties and Hounsfield Number to that of the surrounding soft
tissues.
The X-ray transmission through the TMM compared closely with that
through the water phantom (< 10% difference between the transmission through both
materials, except at very low energies at 50 kVp where the difference was ~ 15%). T1
310
values for the soft tissue surrounding the RA range from 382 ± 13 ms for fat to 1545 ±
142 ms for the kidney medulla [16], so the values measured for the TMM in the
current study (1504 ± 10 ms) are at the upper bounds of this range. The T2 value of
the TMM (40.0 ± 0.4 ms), on the other hand, was fairly typical of those values
occurring for the soft tissue surrounding the RA, which typically range
315
from 29 ± 4 ms for muscle to 81 ± 8 ms for the kidney medulla [16].
Table 2 Comparison of average in vivo soft tissue and measured TMM properties
Parameter
Average soft tissue
in vivo values
1540
0.5 – 0.7
382 – 1545
29 – 81
20 – 70
Speed of sound (m s-1)
Attenuation coefficient (dB cm-1 MHz-1)
T1 (ms)
T2 (ms)
Hounsfield Number
TMM Measured
values
1549 ± 5
0.52 ± 0.03
1504 ± 10
40.0 ± 0.4
58.6 ± 7.3
Visualization of the fiducial markers
320
Images of the fiducial markers obtained with the four imaging modalities are
presented in Figure 6. In each case, the fiducial markers were clearly identifiable and
did not introduce artifacts into the images. For US (Figure 3(a)) the glass beads
appeared as bright spots in the B-mode images, as the glass had a much greater
acoustic impedance (z = 12.1 x 106 kg m-2 s-1) [17] compared to the surrounding
325
TMM (z = 1.6 x 106 kg m-2 s-1) (calculated acoustic impedance (z) = ρc where
ρ density = 1049 kg m-3 and c = 1549 m s-1 [12]). In MRI, the contrast in the
16
T2-weighted images stems from the T2 relaxation times of the phantom constituents
(TMM and glass beads): since the glass beads have an extremely short T2 value
(< 100 s), they effectively produce no signal in a T2-weighted image and
330
consequently appear dark, as shown in Figure 3(b). For the imaging techniques based
on X-ray transmission (DSA and CT), the image contrast stems from a difference in
the X-ray linear attenuation coefficients of the different phantom constituents.
335
Figure 6 Fiducial markers are clearly identifiable in all the imaging modalities (a) ultrasound,
transverse image showing cross section of the lumen and 2 beads at depth 49 mm and 57 mm
on the right side of the phantom, (b) MRI showing a slice with 2 rows of beads, (c)
angiographic image and (d) CT showing a slice with 2 rows of beads
340
17
Visualization of renal artery in phantoms
The images obtained using each imaging modality for the set of phantoms are
345
presented in Figures 7 – 9. Figure 7 shows the 2D B-mode longitudinal US images of
the inlet of each phantom. An example of the 2D transverse B-mode images captured
for three slices through the phantom with a 50% stenosis is shown in Figure 8.
Presented in Figure 9 are the images of the 5 multi-modality phantoms with varying
degrees of stenosis acquired using MRI, CT and angiography contrast images. As can
350
be seen from these images, the vessel of each phantom is clearly visualized and
identifiable.
18
Stenosis
Stenosis
Stenosis
Stenosis
Figure 7 B-mode images showing inlet of the renal artery flow phantom with varying degrees
355
of stenosis (0, 30, 50, 70 and 85%). Reproduced with permission [15]
Stenosis
Figure 8 Transverse views of a phantom with a 50% stenosis at the position of (a) 10 mm
360
after the inlet, (b) the stenosis and (c) 10 mm before the bend.
19
365
MRI
0%
30%
50%
70%
85%
30%
50%
70%
85%
30%
50%
70%
85%
CT
0%
370
DSA
0%
Figure 9 MRI, CT and unsubtracted angiography contrast images of the range of multimodality phantoms
375
Diameter measurements
Previous experiments with the production of the metal models used to create the
lumens in the phantoms demonstrated the reproducibility of the technique, with little
variation in model diameters [15, 18]. The diameter of the vessel in each phantom
was measured from images acquired from each imaging modality. For ultrasound, the
20
380
diameters were calculated from the 2D transverse images such as those shown in
Figure 8. For both MRI and CT, the diameter was calculated from single slices
extracted from their acquired volumes at the mid-point of the vessel, examples of
which are shown in Figure 9.
For DSA, the diameter was calculated from the
subtracted contrast images for each phantom, again as shown in Figure 9.
385
In accordance with clinical practice, based on the dimensions taken from the
2D projection, the degree of stenosis in an artery is defined as;
1  L R100
Equation 1
where L is the diameter of the lumen at the site of the stenosis and R the reference site
diameter, representing the diameter of the lumen pre-stenosis. Using this formula the
390
percentage stenosis was calculated for the reference and for each imaging technique,
where L was taken to be the diameter at location B (at the stenosis) and R the average
diameter from locations A (10 mm before the stenosis) and C (10) mm after the
stenosis. From these measurements, the percentage error for each degree of stenosis
was calculated for each imaging modality. A positive error corresponds to an
395
overestimation in the measurement of the diameter and a negative value an
underestimation error in the measurement of the diameter. A summary of the
over/under estimation of the degree of stenosis (%) with US, MRI, CT and DSA is
presented in Table 3.
400
21
405
Table 3 Summary of over/under estimation of stenosis (%) with US, MRI, CT and DSA
0%
30%
50%
70%
85%
US
0
-7*
-2
-10*
-12*
MRI
2
-3
-3
-5*
-10*
CT
3
-6*
-2
-10*
-13*
DSA
6*
3
-7*
-12*
-14*
P value 0.0045 <0.0001 0.0110 <0.0001 <0.0001
* denotes statistically significant error in the estimation
DISCUSSION
410
The use of rapid prototyping techniques in the fabrication of the anatomically-realistic
renal arteries significantly reduces the time and cost involved in making a range of
phantoms replicating different grades of RAS since only one master silicon mould is
required, which can be reused repeatedly to produce various grades of stenosis with
415
the use of the variable diameter inserts. In previous studies, separate master moulds
were required to produce the different grades of stenosis in each individual phantom
[19, 20]. The TMM chosen for the phantoms was developed as part of the European
Commission-funded project [12] and is the IEC-recommended TMM for use in
ultrasound test phantoms [21]. This TMM has been reported to have acoustic
420
properties similar to soft tissue over the frequency range 2- 20 MHz [12, 14, 22]. To
the authors’ knowledge, this TMM has not previously been used for modalities other
than ultrasound, and thus it is of interest to characterize the MR and X-ray imaging
properties, particularly the Hounsfield Number of this material in order to determine
its suitability across the imaging modalities. The fact that these properties replicated
425
closely those found in vivo for soft tissues in the vicinity of the renal artery gives
confidence in its use as a multi-modality TMM. The T1 relaxation value, which was
at the upper bounds of that found in vivo, could be easily reduced through the addition
22
of a paramagnetic relaxation agent such as MnCl2, which is common practice in the
production of phantoms for MR applications. The X-ray attenuation of the TMM was
430
not determined as part of this study as a calibrated monoenergetic X-ray source was
not available to carry out this measurement, however, the Hounsfield Number of the
TMM was determined and found to be similar to the average value for soft tissue.
All of the renal artery phantoms were successfully imaged with the four imaging
435
modalities, demonstrating the potential of this phantom as a standard multi-modality
vascular test tool. Furthermore, the glass beads used as fiducial markers were clearly
identifiable and did not introduce artifacts into the images. The measured diameters
of the imaged stenoses were determined using a relatively simple 2D approach and
compared to the known values, with the most accurate technique found to be MRI,
440
followed by US, CT and DSA. The reader is referred to Hoffmann et al. for a
discussion of more complex methods for determining vessel diameters in DSA
images, which has been found to improve the accuracy of this technique [23]. There
was a general trend for the modalities to perform worse as the degree of stenosis
increased.
445
For the higher stenosed phantoms (70% and 85%), all four imaging
modalities significantly underestimated the stenosis, with DSA having the greatest
underestimation. The accuracy of US for stenoses < 50%, with mean underestimation
of 3.5% was comparable to results reported in the literature [24, 25].
It has been
shown previously that DSA tends to underestimate diameters [25] and this may be
because it uses a limited number of projections and can therefore fail to detect the
450
maximum stenosis in the artery. Unlike DSA, a 3D volume acquisition technique was
used for the MRI data, while CT acquires projections at different angles, followed by
filtered backprojection image reconstruction, which reduces the partial volume effects
23
compared to the simple DSA approach which can contain pixels with a mixture of
tissue attenuations. The image processing technique of maximum intensity projection
455
which is commonly used in CT produces an overview of the target vessel which aids
the detection of the maximum extent of the stenosis in the vessel. MRI was the most
accurate imaging modality with a mean underestimation of 4%, while both US and
CT had comparable results.
460
The aim of the current study was to demonstrate the potential of the anatomicallyrealistic renal artery phantoms for comparative studies involving US, MRI, CT and
DSA, and therefore default image acquisition parameters were used rather than
making use of pre-set values typically used for specific anatomical regions, which
might have proved inappropriate for the relatively small phantom used in the current
465
study.
Clinically, Spectral Doppler is the most common ultrasound mode for
calculating the degree of renal artery stenosis based on velocity measurements in the
vicinity of the stenosis, while MRA and CTA are increasingly being used to image
renal artery stenosis. The phantoms were primarily designed as flow phantoms, and
further investigation will focus on their use in this capacity with clinical protocols
470
using Spectral Doppler US, MR Angiography (MRA) and CT Angiography (CTA)
techniques.
Although the current study allowed for a standardized comparison between the four
imaging techniques to be carried out in a highly controlled manner, nevertheless a
475
limitation was the use of relatively simple symmetrical stenoses of the same length.
Clinically, stenoses are more complex and present in a variety of lengths and
eccentricities, and furthermore contain fatty and/or calcified material deposited in an
24
asymmetrical manner in the vessel wall and within the lumen itself. Further work by
the authors is focusing on the development of more complex phantoms incorporating
480
a vessel wall mimic, together asymmetrical stenoses incorporating fatty-mimicking
and calcified material, and a pump system producing a pulsatile flow to allow for the
study of the haemodynamic properties of the renal artery system using the different
imaging modalities.
485
CONCLUSIONS
A range of multi-modality anatomically-realistic renal artery phantoms were
490
developed and artifact-free images of the phantoms were obtained using the four
techniques currently used to detect renal artery stenosis (US, MRI, CT and X-ray
DSA). The TMM used in the phantoms’ construction was found to have US, MR and
CT properties comparable to in vivo soft tissue, while the phantoms’ layout proved
suitable for imaging with the various modalities. The severity of stenosis was
495
underestimated in all cases, with the degree of underestimation greatest for the 70%
and 85% stenosed phantoms, with DSA performing consistently worse than the other
imaging modalities.
This study has demonstrated the potential of these novel
phantoms to be used as a standard multi-modality vascular phantom for the evaluation
of RAS-diagnosis techniques,
500
ACKNOWLEDGEMENTS
505
This work was supported by the Technological Sector Research Strand 1 Scheme,
HEA and the TERs 2005 and CABS 2007 Schemes, Research Support Unit of the
Dublin Institute of Technology. C. M. Moran would like to acknowledge funding
from the British Heart Foundation grant PG/07/107/23895.
25
REFERENCES
1.
510
2.
3.
515
4.
520
5.
6.
7.
525
8.
9.
530
10.
535
11.
12.
540
13.
545
14.
15.
16.
550
17.
18.
555
Bosmans, J.L. and M.E. De Broe, Renovascular hypertension: diagnostic and
therapeutic challenges. JBR-BTR, 2004. 87(1): p. 32-5.
Zalunardo, N. and K.R. Tuttle, Atherosclerotic renal artery stenosis: current
status and future directions. Curr Opin Nephrol Hypertens, 2004. 13(6): p.
613-21.
Olin, J.W., et al., The utility of duplex ultrasound scanning of the renal
arteries for diagnosing significant renal artery stenosis. Ann.Intern.Med.,
1995. 122(11): p. 833-838.
Radermacher, J. and R. Brunkhorst, Diagnosis and treatment of renovascular
stenosis--a cost-benefit analysis. Nephrol.Dial.Transplant., 1998. 13(11): p.
2761-2767.
Reynolds, H.R., et al., Frequency of severe renal artery stenosis in patients
with severe thoracic aortic plaque. Am J Cardiol, 2004. 94(6): p. 844-6.
Smith, R.F., et al., Geometric characterization of stenosed human carotid
arteries. Acad Radiol, 1996. 3(11): p. 898-911.
Cloutier, G., et al., A multimodality vascular imaging phantom with fiducial
markers visible in DSA, CTA, MRA, and ultrasound. Med Phys, 2004. 31(6):
p. 1424-33.
Dabrowski, W., et al., A real vessel phantom for imaging experimentation.
Med Phys, 1997. 24(5): p. 687-93.
Frayne, R., et al., A geometrically accurate vascular phantom for comparative
studies of x-ray, ultrasound, and magnetic resonance vascular imaging:
construction and geometrical verification. Med Phys, 1993. 20(2 Pt 1): p. 41525.
Meagher, S., et al., Anatomical flow phantoms of the nonplanar carotid
bifurcation, part II: experimental validation with Doppler ultrasound.
Ultrasound Med Biol, 2007. 33(2): p. 303-10.
Allard, L., et al., Multimodality vascular imaging phantoms: a new material
for the fabrication of realistic 3D vessel geometries. Med Phys, 2009. 36(8): p.
3758-63.
Teirlinck, C.J., et al., Development of an example flow test object and
comparison of five of these test objects, constructed in various laboratories.
Ultrasonics, 1998. 36(1-5): p. 653-60.
Ramnarine, K.V., T. Anderson, and P.R. Hoskins, Construction and geometric
stability of physiological flow rate wall-less stenosis phantoms. Ultrasound
Med Biol, 2001. 27(2): p. 245-50.
Browne, J.E., et al., Assessment of the acoustic properties of common tissuemimicking test phantoms. Ultrasound Med Biol, 2003. 29(7): p. 1053-60.
King, D.M., et al., Development of a range of anatomically realistic renal
artery flow phantoms. Ultrasound Med Biol, 2010. 36(7): p. 1135-44.
de Bazelaire, C.M., et al., MR imaging relaxation times of abdominal and
pelvic tissues measured in vivo at 3.0 T: preliminary results. Radiology, 2004.
230(3): p. 652-9.
Selfridge, A.R., Approximate material properties in isotropic materials. IEEE
Trans Sonics Ultrasound, 1985. SU-32: p. 13.
King, D.M., Development of renal phantoms for the evaluation of current and
emerging ultrasound technology, in School of Physics. 2009, Dublin Institute
of Technology: Dublin.
26
19.
560
20.
21.
565
22.
23.
570
24.
25.
Poepping, T.L., et al., An in vitro system for Doppler ultrasound flow studies
in the stenosed carotid artery bifurcation. Ultrasound Med Biol, 2002. 28(4):
p. 495-506.
Watts, D.M., et al., Anatomical flow phantoms of the nonplanar carotid
bifurcation, part I: computer-aided design and fabrication. Ultrasound Med
Biol, 2007. 33(2): p. 296-302.
IEC, Ultrasonics - Flow measurement systems: Flow test object, G.
International Electrotechnical Commission, Switzerland, Editor. 2001.
Brewin, M.P., et al., The acoustic properties, centered on 20 MHZ, of an IEC
agar-based tissue-mimicking material and its temperature, frequency and age
dependence. Ultrasound Med Biol, 2008. 34(8): p. 1292-306.
Hoffmann, K.R., et al., Vessel size measurements in angiograms: a
comparison of techniques. Med Phys, 2002. 29(7): p. 1622-33.
Beux, F., et al., Automatic evaluation of arterial diameter variation from
vascular echographic images. Ultrasound Med Biol, 2001. 27(12): p. 1621-9.
Boussion, N., et al., Geometrical accuracy and fusion of multimodal vascular
images: a phantom study. Med Phys, 2004. 31(6): p. 1434-43.
575
27